Accounting SSD wear to vservers: First we must determine the device number to name mapping: /sys/block/ /dev determines the major minor /device -> symlink to real device . If no symlink it's no real device. /device/model -> scsi id
filter on model or device name? smsbeest:/sys/devices/pci0000:00/0000:00:1f.2# ls -al ata*/host*/target
SSD accounting needs the total write stats. There are two write stats available: smsbeest:/dev/cgroup/obproxy3# grep Write blkio.throttle.io_service_bytes blkio.io_service_bytes blkio.throttle.io_service_bytes:8:48 Write 7059456 blkio.throttle.io_service_bytes:8:0 Write 7059456 blkio.throttle.io_service_bytes:9:4 Write 7061504 blkio.io_service_bytes:8:48 Write 6750208 blkio.io_service_bytes:8:0 Write 6750208
I am not really sure what's the difference between the two and why the numbers differ.
Modular point is: proc file parser that does something like: read line, split line into Maybe a generic parser like: <key n+1> which returns a value, when there is a single value, a key n -> value when there is just that, or a key n+1->key n->value